Repolode
Submit
yokingma

yokingma/SearChat

AI-powered conversational search engine that answers questions with live web results, supporting OpenAI, Anthropic, Gemini and Vertex AI models.

Open-source alternative to:Perplexity

View on GitHubWebsite
Stars
1,062
Forks
182
Open issues
2
Last commit
2 months ago

What is it?

SearChat is a conversational search engine: you ask a question in a chat interface, the system runs web searches, and an LLM writes an answer based on the results. It is built as a Turborepo monorepo with a Koa and TypeScript backend and a Vue 3 frontend.

The project connects to OpenAI-compatible, Anthropic, Google Gemini and Google Vertex AI model APIs, configured through a model.json file. Search can run through SearXNG, which needs no API key, or through Bing, Google, Tavily, Exa, Bocha and ZhipuAI's search plugin. A Deep Research mode runs iterative searches and produces a structured report with citations, built on LangChain and LangGraph.

The repository includes a Docker Compose deployment that starts the app together with SearXNG, and it also documents a standalone deepsearcher npm package for embedding Deep Research in Node.js projects.

Key features

  • 01Chat-style multi-turn search with streamed answers
  • 02Supports OpenAI-compatible, Anthropic, Gemini and Vertex AI models
  • 03Multiple search backends including SearXNG, Bing, Google, Tavily and Exa
  • 04Deep Research mode with iterative searches and cited reports
  • 05One-command Docker Compose deployment with SearXNG included

Use cases

  • Self-hosting a private AI search assistant with your own model API keys
  • Running a free-search setup via SearXNG without paid search API keys
  • Adding Deep Research capabilities to a Node.js application through the deepsearcher package

Quick start

  1. 1

    Install Docker and Docker Compose

  2. 2

    Create docker-compose.yaml and model.json, then configure API keys

  3. 3

    Start services

    docker compose up -d
  4. 4

    Open http://localhost:3000 in your browser

Things to know

  • The model must support Tool Call (Function Calling) to get the best results.
  • Search engines such as Bing, Google, Tavily, Exa and Bocha require their own API keys, while SearXNG needs Google network access.

In collections

Claude Code tools, skills & pluginsSelf-hosted apps

Topics

#ai-search#llm#rag#deep-research#searxng#self-hosted#docker#typescript#ai#anthropic

FAQ

What is SearChat?
AI-powered conversational search engine that answers questions with live web results, supporting OpenAI, Anthropic, Gemini and Vertex AI models. SearChat is a conversational search engine: you ask a question in a chat interface, the system runs web searches, and an LLM writes an answer based on the results. It is built as a Turborepo monorepo with a Koa and TypeScript backend and a Vue 3 frontend.
Is SearChat free to use? What license does it have?
SearChat is open source under the MIT license. Check the license terms before using it in commercial products.
Is SearChat still maintained?
Updated within 6 months. The last commit to SearChat was on Sep 1, 2026.
What can I use SearChat for?
Self-hosting a private AI search assistant with your own model API keys Running a free-search setup via SearXNG without paid search API keys Adding Deep Research capabilities to a Node.js application through the deepsearcher package
How do I install SearChat?
Install Docker and Docker Compose Create docker-compose.yaml and model.json, then configure API keys Start services: docker compose up -d Open http://localhost:3000 in your browser
Where can I get SearChat?
The source code and installation instructions are on GitHub: https://github.com/yokingma/SearChat

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects